Here is the "Micro" folio. Best (or "most lucky") performers were MGC and MSF. Most unlucky was M6B. In a next post will show some trades in detail. [trading contains also rollover (except MGC), executed yesterday to test the functionality]. This folio has little DD and little PNL, and is probably more suitable for people with relatively small capital or just starting out with trading robots (the whole game is scaled down about 10 times). <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3118549" width="1200"> Tom
These are, for instance, the trades of MSF (the 2-nd "most lucky", after MGC). The maroon vertical line indicates the rollover instant. I noticed that for all micro instruments (except M6A) the old and the new instrument prices were very close, so that the 2 price curves joined quite smoothly. <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3118663" width="1200" /> Tom
The micro folio just closed all positions. MJY roared up (probably due to the earthquake (?) ). <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3119275" width="1200" /> Tom
I bet you are curious to see what looked like MJY spike up ) The picture shows also the <b>"extreme entry bounds"</b> (the 2 horizontal lines, blue and red) established by the bot while running (clearly, now the user can define initial "discretionary" bounds). <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3119410" width="1200"> (the vertical line, as said, is the rollover instant) Tom
It seems that working with entry bounds is greatly beneficial to autotrading. After all this corresponds to an intuitive notion that many <b>successfull discretionary traders </b>are applying everyday. The <b>"buy the dips, sell the rips"</b> wall street adage effectively syntetizes this behavior. And the bot can do that for us, by automatically determining (or a large number of instruments) trading range and "best" places for entries (an activity that a discretionary trader can do, but with some work and effort). Clearly, even buying a dip one has no (absolute) guarantees, but that may be better than systematically buying on an historical top price! In the long run, this systematic behavior can save from several deep moves against the initial position where the robot has to engage in smart protective mechanisms, often with some drawdown or wasted money due to stops/reversions on local fluctuations. An this behavior, applied systematically, can greatly improve overall performances (especially in terms of the ratio $ / Risk). Any thoughs ? Tom
I have restarted the mini folio with the latest strategy. I have anticipated the stop/reverse to 4, which is still plenty of space to capture large fluctuations. Captured 7K in few hours. Mostly due to CL, HG, SI. <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3121664" width="1200"> In addition to discretionary bounds, that are at instrument level, I have also introduced the possibility to define "zones" to bound entries which make more flexible the constraint of "extreme entry". I am trying to setup strategies with earlier reversal, but so far simulations indicated a decrease of performances ($/Risk). Will continue the search ... Tom
As curiousity, here are the trades of the 2-nd most lucky today. It's CL (after SI). <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3122033" width="1200" />
Tom, Can you clairfy your backtesting procedure a little please. This thread is quite long and I might have missed it, but do you show a long-term backtest (a year plus) and can you tell me when the parameter selection (optimisation) was last made in that period. Moon
Welcome Moon, actually <a href="http://www.datatime.eu/public/gbot/Strats%20G-BOT/Strategy_CT_T_LS/Strategy_CT_T_LS.htm">that</a> is equivalent to 71 years ;-)) You may have been missed that part of the thread. We have been discussing simulation models using Geometric Brownian motion, in order to have sufficient "diversity" in the tickdata to avoid curve fitting. The entire simulation is done in this case with a given rule set, which you can read on the result page. Clearly, it can be changed. I am always trying to improve performances with new ideas. Tom
Hi Tom, That's very interesting indeed. Do you have any data for Ex-Post returns - that's what I was looking for.